As photolithographic techniques have succeeded in producing ever smaller feature sizes, the demands on the performance of the lithographic imaging system have become increasingly stringent. In particular, the intrinsic ability of the photoresist film, which may be regarded as an element in the imaging system, to resolve small features has become a matter of great concern, since the inherent resolution of a photoresist places a limit on the dimension of the smallest feature that can be produced. Accordingly, the quantitative characterization of a photoresist's spatial response is important for resolution-enhancing technologies such as optical proximity correction (OPC), phase-shifting masks, and modified (or off-axis) illumination, all of which are commonly used in state-of-the-art lithography.
The performance of a photoresist depends not only on its chemical composition, but also on its processing conditions (e.g., the pre- and post-expose bake temperatures and duration, the development time and temperature, etc.) and the environment in which it is used. Because any and all of the conditions that might affect the performance of a photoresist are potentially variable, it is valuable to objectively and quantitatively characterize the spatial resolution of a photoresist film under the photoresist's preferred processing conditions. Ideally, such a characterization should be performed efficiently, i.e., with sufficient speed and ease of use that it can be applied routinely in a realistic manufacturing environment. In practical terms, this necessitates that such a technique be compatible with conventional lithographic tools and procedures.
To this end, a method was recently demonstrated for accurately characterizing the spatial resolution of photoresist materials (see J. A. Hoffnagle, W. D. Hinsberg, M. I. Sanchez, and F. A. Houle, Optics Letters, vol. 27, pp. 1776–1779, 2002). A key step in this methodology is the exposure of the photoresist film to the sinusoidal optical intensity pattern produced by the interference of two coherent, plane-wave beams of wavelength λ, each of which intercepts the film at an angle θ (with respect to the normal to the film). This step produces a “grating pattern” in the photoresist film that has a period Λ=λ/2 sin θ and an intensity distribution given byI(x)=Dq[1+cos(Kx)]/2  (1)in which D is the dose or total beam energy used to write the grating pattern, q is a coefficient that relates dose to absorbed energy per unit area (and therefore depends on both the beam cross section as well as the absorbitivity and thickness of the film), K=2π/Λ is the wave vector of the grating pattern, and x denotes a particular spatial coordinate.
One then assumes that the developable latent image, taken here to be the density ρ(x) of the photoresist polymer that is modified as a result of this exposure, is simply equal to the convolution of the optical illumination pattern and the point-spread function (PSF) of the photoresist. If f(x) represents the line-spread function (LSF), i.e., the one-dimensional analog of the PSF, then it can be shown thatρ(x)=p Dq[1+a(K)cos(Kx)]/2  (2)in which p is a proportionality constant that relates absorbed optical energy density to the density of modified polymer, and a(K) is the modulation transfer function (MTF), which is equal to the Fourier cosine transform of the line-spread function f(x).
Although the latent image in the photoresist film is not generally directly observable, it is possible to determine the MTF (and thus the LSF) from the variation of the developed linewidth with exposure, given a quantitative model of resist development. In the case of a simple threshold model, in which there is a threshold dose below which the photoresist film is not developed, it can be shown thatD/D0=1/[1+a(K)cos(Kt)]/2  (3)in which t is the developed linewidth, and D0 is the dose that would be necessary to develop the entire exposed region. Thus, it is possible to determine D0 and a(K) for a given beam wavelength λ by fitting measurements of developed linewidth t and dose D to equation (3). By repeating this analysis over a range of different incident spatial frequencies, the MTF a(K) can be determined over that range. From the MTF, it is possible to determine the LSF f(x), which is essentially a measure of the photoresist's inherent spatial resolution at a particular wavelength.
Unfortunately, the means used to generate a sinusoidal optical grating in a photoresist film can not be easily integrated with standard photolithographic tools, and this incompatibility greatly limits the usefulness of the method of Hoffnagle et al. described above. The currently preferred method for generating a sinusoidal optical grating relies on the interference of two mutually coherent light beams. If two beams of wavelength λ illuminate a surface, and each beam makes an angle θ with respect to the normal to that surface, then the angle between the two beams is 2θ and the resulting interference pattern is sinusoidal withK=4π sin(θ)/λ  (4)
Many interferometers have been designed and built that generate the two required interfering beams, but these are generally special-purpose instruments that must be carefully aligned for each particular value of K (spatial frequency) that is desired. The need to realign an interferometer for each spatial frequency greatly slows the process of generating sets of gratings spanning a large spatial frequency range, which is needed to carry out the procedure described above. Such interferometers have little in common with present-day lithographic exposure tools, which use projection optics to illuminate a reticle and transfer the image of the reticle to a surface.
The method most commonly used for generating a sinusoidal optical grating with a lithographic exposure tool is the non-interferometric one of preparing a reticle on which the transmission of the illuminating light varies sinusoidally. This approach is impractical for high spatial frequencies and short wavelengths, because of the difficulty of accurately patterning the reticle substrate with a semitransparent material having precisely controlled optical density.
Consequently, there is at present no suitable method for generating sinusoidal optical gratings using conventional lithographic tools. If existing projection exposure tools could be adapted to generate accurate sinusoidal optical gratings at the surface of a photosensitive material in an efficient manner, the methodology of Hoffnagle et al. could be used routinely to characterize the spatial response of the resist materials, which would be of great value for lithography. What is needed therefore, is a technique for integrating interferometric optics with the optics of projection lithography. The present invention satisfies this need.